Evaluating auto-translation quality: a case study of YouTube

• Main Author: Chen, Hanzhang
• Other Authors: -
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Arts and Humanities; YouTube; Machine translation; Auto-translation; Translation errors; Gemini
• Online Access: https://hdl.handle.net/10356/178386
• Institution: Nanyang Technological University

This study examines the quality of YouTube's auto-translation English-to-Chinese subtitles. The research explores 15 videos across different genres and found 84 major translation errors which could be divided into 7 categories based on the framework of the Multidimensional Quality Metrics (MQM) framework. Moreover, the study delves into the translation capability of Gemini and compares the errors generated by Gemini with those generated by the YouTube auto-translation system. The study finds that Gemini could significantly improve the translation quality.